Abstract
Background: With respect to sporadic Creutzfeldt–Jakob disease (sCJD), six molecular subtypes (MM1, MM2, MV1, MV2, VV1, and VV2) have been described, which vary with respect to age at disease onset, disease duration, early symptoms, and neuropathology. MRI signal alterations were reported to correlate with distinct Creutzfeldt–Jakob disease (CJD) subtypes. This multicenter, international study aimed to describe the brain MRI findings associated with each of the sCJD molecular subtypes.
Methods: Pathologically confirmed sCJD cases with codon 129 genotype (MM, MV, and VV), PrPSc type, and fluid-attenuated inversion recovery (FLAIR) or diffusion-weighted imaging (DWI) were collected in seven countries. All MRI scans were assessed for signal changes according to a standard protocol encompassing seven cortical regions, basal ganglia, thalamus, and cerebellum.
Results: MRI scans were evaluated in 211 CJD patients (98 MM1, 23 MM2, 19 MV1, 30 MV2, 9 VV1, and 32 VV2). Basal ganglia hyperintensities occurred most frequently in MV2, VV2, and MM1 subtypes (79, 77, and 70%). Wide cerebral cortical signal increase was most common in VV1, MM2, and MV1 subtypes (86, 77, and 77%). Thalamic hyperintensities occurred most often in VV2 (45%) and MV2 (43%). The most consistent finding across most subtypes was high signal in basal ganglia, with these abnormalities found in 63% (FLAIR) and 71% (DWI).
Conclusion: Cortical signal increase and hyperintensities in the basal ganglia and thalamus are detected by MRI across all molecular sporadic Creutzfeldt–Jakob disease subtypes. Our findings argue that characteristic MRI lesion patterns may occur for each molecular subtype.
Glossary
- AL=
- anterolateral;
- AUC=
- area under the receiver operating characteristic curve;
- BG=
- basal ganglia;
- CI=
- confidence interval;
- CJD=
- Creutzfeldt–Jakob disease;
- DWI=
- diffusion-weighted imaging;
- FLAIR=
- fluid-attenuated inversion recovery;
- MD=
- mediodorsal;
- OR=
- odds ratio;
- ROC=
- receiver operating characteristic;
- sCJD=
- sporadic Creutzfeldt–Jakob disease.
MRI has played a role in the diagnosis of sporadic Creutzfeldt–Jakob disease (sCJD) for many years, and various signal abnormalities have been reported.1–5
Heterogeneity in sCJD correlates with the codon 129 genotype of the prion protein gene (PRNP) in combination with the existence either of two distinct types of pathologic prion protein (PrPSc 1 or 2). Based on these variables, six sCJD subtypes (MM1, MM2, MV1, MV2, VV1, and VV2) have been defined, and variations in the sensitivities of diagnostic tests for the different subtypes have been reported.6–11
Although there is now considerable experience with the use of MRI in sCJD, findings in relation to the specific subtypes are more limited.5,9,12–15 The general conclusion from these data are that characteristic MRI lesion patterns may correspond to a specific Creutzfeldt–Jakob disease (CJD) subtype.
To address this issue, we undertook a multicenter, international collaborative study to describe the MRI lesion patterns across the entire clinical spectrum of sCJD. All subjects included in the study had undergone MRI using the most sensitive pulse sequences, with systematic regional brain assessment for abnormal signal changes.
METHODS
Patients.
Cases were included from seven countries (United Kingdom, France, Italy, Belgium, Germany, Canada, and Australia) according to the following criteria:
-
CJD diagnosis confirmed by brain pathology (definite cases)
-
Molecular subtype determined by codon 129 genotyping (MM, MV, or VV) and Western blot analysis of brain PrPSc type (1 or 2) (corresponding to MM1, MM2, MV1, MV2, VV1, or VV2 subtype) and3 fluid-attenuated inversion recovery (FLAIR) or diffusion-weighted imaging (DWI) MRI of the brain available
All three criteria had to be fulfilled in each case. Genetic CJD (causal mutations found in PRNP), iatrogenic, and variant CJD cases were excluded.
Information on CSF (14-3-3 protein) and EEG (periodic sharp wave complex) findings was acquired and data are displayed in table e-1 on the Neurology® Web site at www.neurology.org. EEGs were reviewed by a member of the national surveillance and scored positive according to the criteria.16 The CSF 14-3-3 immunoassays were performed using Western blotting (conformity of testing methods and results interpretation confirmed by blinded sample exchange program as reported previously).17
The study comprised 211 patients who died between March 1996 and February 2006: 92 (44%) from Germany, 38 (18%) from the United Kingdom, 34 (16%) from Italy, 17 (8%) from Australia, 15 (7%) from Belgium, 9 (4%) from Canada, and 6 (3%) from France. In 8 patients, mixed PrPSc types were detected, and they were excluded from the article. According to the study goal of characterizing the single subtypes by MRI, the inclusion of MM1 patients as most frequent molecular type was stopped toward the end of the study because of sufficient patient numbers.
Molecular subtype classification was performed as published previously,6 determined by the combination of the codon 129 genotype of the PRNP (MM, MV, or VV) and the pathologic isotype of the prion protein (PrPSc 1 or 2).
The PrPSc typing was performed according to standard methods.7 The number of examined brain regions usually included the cerebral cortex, basal ganglia or thalamus, and cerebellum.8 Codon 129 status was determined either as part of genotyping of the entire PRNP open reading frame or by restriction fragment length polymorphism analysis.8
MRI data.
The magnetic resonance images were acquired at local sites and obtained from scanners of different manufacturers and of different magnetic field strengths of 0.5 to 3 tesla (mainly 1.5 tesla).
The majority of the scans were available as hard copy. If serial MRIs were available, the first examination was used for the analysis. The scans were assessed for hyperintense signal abnormalities by neuroradiologists (K.K., P.S., and D.C.) aware of the CJD suspicion but not aware of the molecular subtype. The scans were reviewed by each radiologist individually. All scans were assessed by a neuroradiologist (K. K.) aware of the suspicion of sCJD but not aware of the diagnosis or the molecular subtype.
Two further neuroradiologists (P.S. and D.C.) assessed two separate MRI series for the estimation of the interobserver agreement. With the three raters' results, we calculated 1) the percentage of concordant results and 2) the κ statistic, which compares the agreement against that which might be expected by chance. Interobserver agreement was moderate between raters 1 and 2 (concordance 76%, κ = 0.45, p < 0.001) and high between raters 1 and 3 (concordance 82%, κ = 0.62, p < 0.001). In rare cases with discrepant findings, a consensus between three radiologists was sought. A standardized protocol including seven cerebral cortex regions (frontal, parietal, temporal, occipital, cingulate gyrus, insula, and hippocampus), basal ganglia (caudate nucleus, putamen, and globus pallidus), thalamus (anterolateral nuclei, mediodorsal nuclei, and pulvinar), and cerebellar cortex was used.18 For the thalamus, the presence of a pulvinar sign or hockey stick sign was also rated.19 The quality of the complete MRI examination was graded from 1 to 6 (1 = excellent, 6 = poor). Scans graded higher than 4 were excluded from the study (n = 6); no later scans were available in these patients.
Statistical analysis.
To describe different sCJD MRI patterns, we aimed to select the best radiologic features for molecular subtype discrimination. We considered a brain region as affected when a high signal was found in either FLAIR or DWI MRI sequences.
General data analysis.
For the radiologic features selected, we compared the proportion of patients with a high signal in each studied region against that same proportion in the rest of patients not belonging to that particular subtype, and so on for the six subtypes. Therefore, to test how well the radiologic feature discriminated each particular subtype, we used six dichotomous outputs (MM1 vs all others, MM2 vs all others, MV1 vs all others, MV2 vs all others, VV1 vs all others, and VV2 vs all others) to fit six binary logistic regression models. Odds ratios (ORs) were calculated from the logistic regression analyses. First, we performed univariate or crude analysis showing the increase or decrease in odds of the subtype-specific diagnosis when each of the radiologic signs were present. Secondly, we performed hierarchical cluster analysis to gain more insight about the relationships among the different radiologic signs. We specifically were interested to know which radiologic signs grouped together and therefore tended to appear simultaneously in the same patients. The hierarchical cluster analysis is an exploratory procedure useful for finding natural groupings and discovering hidden structures in data. The basic criterion for any clustering is distance. Objects that are near each other should belong to the same cluster, and objects that are far from each other should belong to different clusters. We performed cluster analysis using the overall sample of sCJD MRI scans. Because our main interest was to examine relationships between the different radiologic signs, the objects clustered in our analysis were the brain MRI variables (regions assessed in the MRI scans), which were all categorical (high signal [yes/no]), and the study subjects were treated as the variables of the analysis. We selected between-groups average linkage as the clustering method for our analysis, and simple matching coefficient as the measure of distance between variables. The information resulting from the cluster analysis helped us to establish, independently of the molecular subtype of the patients, which signs could be pooled into composite variables; e.g., all thalamic nuclei were in the same cluster, so we subsequently used “any thalamic nuclei affected” as a new variable to test for molecular subtype discrimination (figure e-1). Clusters of variables were pooled to form composite variables when at least one of the variables included was significantly associated with subtype discrimination in the crude analysis.
We chose the following radiologic signs as sCJD subtype predictors: 1) more than three cerebral cortical regions affected, 2) hippocampus affected, 3) any basal ganglia affected, 4) any thalamic nuclei affected, and 5) cerebellum affected. Multivariate analysis including the five selected predictors was performed, for each disease subtype (yes/no) separately, to assess them independently of each other.
PrPSc type predictive model.
We built a PrPSc type predictive model including, together with the MRI data, PRNP codon 129 genotype, and clinical information. The objective of this analysis was to assess which of these variables better predicted the type of PrPSc (1 or 2). We fitted a logistic regression model using the overall sample of sCJD MRI scans. The variable PrPSc type (1or 2) was the model's output, and the same five MRI variables plus age at onset, disease duration, CSF, and EEG results were included as predictors. We also included sex as a covariate. A second PrPSc type predictive model was fitted including codon 129 PRNP genotype information as well. The performance of both models, clinical and clinical plus genetic, were evaluated by receiver operating characteristic (ROC) curves. We calculated the area under the ROC curve by the nonparametric method implemented in SPSS version 15.0 for Windows software.
Ethics.
Approvals by local ethical standard committees were obtained by each national CJD surveillance unit in participating countries.
RESULTS
Patients.
Patient characteristics according to CJD subtype are shown in table e-1.
MRI findings.
For all patients, MRI was performed a median of 3.1 months (range 0–55.3 months) from the onset of symptoms, corresponding to the second third of the whole disease duration in the majority of patients (table e-1). The spectrum of available sequences comprised 174 FLAIR and 113 DWI (distribution of grades given in table e-1).
Basal ganglia signal increase was found in 71% (DWI) and 63% (FLAIR) of all patients. Widespread signal increase of the cerebral cortex (more than three regions affected) was found in 66% (DWI) and 38% (FLAIR) of all patients (table e-1).
MRI findings in sCJD subtypes.
DWI detected a higher percentage of signal alterations than FLAIR. The frequency of hyperintensities observed for each subtype on DWI images is given in table e-2.
For MM homozygotes, the frontal and parietal lobes signal changes were frequent, but in MM2 subjects, the signal increase was more widespread, commonly including the temporal lobes and hippocampus. In MV heterozygotes, the frontal lobes and cingulate gyri were most affected. However, in the MV1 subtype, high rates of signal increase were found in the insular cortex, whereas in MV2 subjects, cerebellar signal increase was observed more frequently in DWI. The cingulate gyri were the most frequently affected cerebral cortex region in the VV2 subtype. In VV1 subjects, DWI studies were limited, but FLAIR revealed a higher frequency of hyperintensity in the parietotemporal lobes and particularly the insular cortex (71% vs 14%) compared with VV2 cases.
Crude analysis and cluster analysis of the raw data.
Based on the ORs and p values obtained by crude analysis (table 1) and hierarchical cluster analysis of MRI findings across the spectrum of subtypes, five MRI criteria were selected as most suitable variables for discrimination between the subtypes:
Table 1 Radiologic predictors of Creutzfeldt–Jakob disease subtypes
-
Cortex: widespread involvement (including more than three cortex regions)
-
Hippocampus region affected
-
Basal ganglia affected (caudate nucleus or putamen)
-
Thalamus affected (any of the three nuclei)
-
Cerebellar cortex affected
The five selected variables were included in a logistic regression model for multivariate analysis. Table 2 shows whether the presence (OR >1.0) or absence (OR <1.0) of the MRI finding was significantly related to one of the subtypes.
Table 2 Multivariate analysis including composite predictors
Detailed analysis of individual subtypes.
MRI examples showing characteristic findings of the subtypes are given in figure 1. The percentages and p values used for the characterization of the subtypes are displayed in tables 1 and 2 and figure 2.
Figure 1 Characteristic MRI findings in sporadic Creutzfeldt–Jakob disease (sCJD) subtypes
Diffusion-weighted images of six sCJD patients with various molecular subtypes showing basal ganglia signal increase and signal increase in the frontal, temporal, and insular cortex (MM1); predominant cortical signal increase in the frontal and parietal lobes (MV1); cortical hyperintensities in the cingulate gyrus, insular cortex, and hippocampus (VV1); basal ganglia and widespread cortical hyperintensities (MM2); predominant signal increase in the basal ganglia and thalamus (MV2); and predominant basal ganglia signal increase and signal increase in the cingulate gyrus (VV2). The MV1 image was published in Am J Neuroradiol 2008;29:1519–1524 (© 2008 American Society of Neuroradiology; reprinted with permission).22 The VV1 image was published in Neurology 2005;65:1544–1550 (© 2005 AAN Enterprises, Inc.; reprinted with permission).15 The MV2 image was published in Am J Neuroradiol 2006;27:1459–1462 (© 2006 American Society of Neuroradiology; reprinted with permission).23
Figure 2 MRI lesion patterns in MM1 and MM2, MV1 and MV2, and VV1 and VV2
*As seen in fluid-attenuated inversion recovery or diffusion-weighted imaging.
MM1.
Basal ganglia signal increase was frequent in the MM1 subtype, and cerebral cortex involvement was widespread in half the patients. The frontal lobes, parietal lobes, and cingulate gyri were most frequently affected (table e-2). The absence of hippocampal and thalamic signal increase differentiated this subtype from others.
MM2.
No patient was classified as MM2-thalamic through neuropathologic examination (in 13 of the 23 MM2 patients, no thalamic and brain stem material was available). Because MM2-thalamic types have been previously described as showing no signal alterations on the MRI,20 we examined the MM2 group for MRI negative cases: all MM2 patients were found to have signal abnormalities on all available sequences, making MM2-thalamic less likely.
Widespread cortical signal increase, which typically included the temporal lobes, was characteristic of this subtype. Basal ganglia involvement was rather limited, and the absence of this finding supported the diagnosis. In contrast to MM1 patients, thalamic signal increase and cerebellar signal increase occurred more frequently in MM2 types.
MV1.
In MV1 patients, cerebral cortex and basal ganglia were both involved often. The cortical signal increase typically included the insula and the hippocampus.
MV2.
The basal ganglia and the thalamus were characteristically affected in the MV2 subtype. Thalamic signal increase was most frequently observed in the pulvinar, followed by the mediodorsal nuclei and the anterolateral nuclei. A pulvinar sign was present in three patients, with a hockey stick sign in one. The cerebral cortex involvement was rather limited and most frequently included the frontal lobes and cingulate gyri.
VV1.
The VV1 subtype showed the most frequent cerebral cortical signal changes, with the most affected cortex region being the cingulate gyri, followed by the insula and the temporal lobes (tables 1 and 2). Basal ganglia or thalamic signal increase was typically absent.
VV2.
Across all molecular subtypes, VV2 patients showed the most frequent involvement of basal ganglia and thalamus. Cerebral cortical signal increase was usually restricted to less than three regions and most frequently found in the cingulate gyrus (tables 1 and 2).
Predictor analysis.
Limited cerebral cortical hyperintensities and the presence of thalamic hyperintensities were significantly related to PrPSc type 2 (table 3) as well as valine homozygosity at codon 129, age at onset, and prolonged disease duration.
Table 3 PrP type predictive model
DISCUSSION
This multicenter collaborative study has determined the brain MRI findings in a large number of cases of sCJD with defined molecular subtype according to the codon 129 genotype and PrPSc type. Although basal ganglia hyperintensities on the MRI represented a consistent finding in all subtypes except VV1, the frequency and location of cortex hyperintensities as well as the presence or absence of thalamus involvement varied between the subtypes.
The most characteristic MRI lesion patterns were found in MV2 and VV2 showing predominant involvement of thalamus and basal ganglia. Limited cortical signal increase was significantly related to PrPSc type 2. A further possible characteristic lesion pattern was found in VV1 showing widespread cortical hyperintensities and absence of basal ganglia signal alterations. In the other subtypes, there was a greater overlap between cortical and subcortical involvement.
In MV2, we found less basal ganglia and thalamic involvement than previously reported.5 This discrepancy may be based on the use of only the first MRI study images in our study and differences in the patient case-mix. The pulvinar sign according to current criteria was identified in the MV2 subtype only.19 Because of the generally high frequency of thalamic hyperintensities in MV2, this subtype is the most likely to be mistaken for variant CJD on MRI.
MRI findings in the VV2 subtype have been reported previously in three cases, demonstrating limited cerebral cortex involvement associated with strong subcortical signal increase.14,21 Consistent with this, we found widespread thalamic involvement in our VV2 patients, with a high degree of the involvement of the mediodorsal and anterolateral thalamic nuclei.
VV1 and MM2-cortical subtypes have been previously described as atypical CJD variants because of longer disease durations with relatively slowly progressive dementia and absence of typical EEG changes. For VV1, basal ganglia hyperintensities are rare,15 and in MM2-cortical, isolated cerebral cortex involvement with limited basal ganglia involvement is characteristic, although normal MRI scans have been reported.12,13 In our study, 10 patients were classified as MM2-cortical, and widespread cortical signal increase represented the main characteristic.
Patients with MM1 and MM2 types, basal ganglia and widespread cortex involvement were found in at least 50% of both patient groups, leading to a high overlap. However, hippocampus involvement was more frequently found in MM2-cortical and was typically absent in MM1.
Previously, the overlap in clinical and neuropathologic findings in some studies has prompted that MM1 and MV1 patients should be combined as one phenotype,6 although significant differences in investigation findings and clinical features have been noted between these two subtypes.8 Our study offers additional evidence that MM1 and MV1 might be considered as separate subtypes, by showing that MRI lesion profiles differ, with MV1 showing more frequent cortical and thalamic involvement.
For the premortem diagnostic evaluation, characteristic brain MRI lesion patterns might be helpful in establishing a diagnosis of sCJD and may help to identify atypical sporadic disease forms.
Summarizing our data, basal ganglia and cortical hyperintensities (limited or widespread) represent the most frequent MRI finding in CJD and are most typically found in MM1 subjects with a rapid disease course but also in MV1 individuals. In MM2, the disease course is more prolonged and widespread cortical hyperintensity on the MRI represents the predominant finding. Finally, predominant subcortical signal increase with limited cortical hyperintensities (mainly in the cingulate gyrus) was seen in the MV2 or VV2 type of CJD.
This is the most comprehensive study on MRI findings in sCJD to date and, in combination with previous studies, provides firm evidence of the high sensitivity of brain MRI in the diagnosis of sCJD. Some hyperintensity patterns, such as involvement of more than three cortical areas and hyperintensities in the basal ganglia, should be further evaluated and discussed as potential parameters for inclusion into diagnostic clinical criteria.
Because we intended to study lesion patterns on the MRI in different disease subtypes, no control group was included. MRI changes are frequent, and knowledge about different patterns may help to recognize specifically atypical disease variants.
Nevertheless, with acknowledgment that MRI changes in sCJD differ across the subtypes and are only a surrogate disease marker, any proposed diagnostic criteria will need to be rigorously validated with respect to sensitivity and specificity in a cohort of patients presenting the differential diagnosis of rapidly progressive dementia.
ACKNOWLEDGMENT
The authors thank the physicians for case notification and provision of clinical and pathologic data and MRI scans; Dr. Piero Parchi, Prof. Salvatore Monaco, Dr. Gianluigi Zanusso, Dr. Carlo Buffa, and Dr. Bergeron for neuropathologic evaluation and PrP Western blot typing; Dr. Sabina Capellari, Dr. Daniele Imperiale, Dr. Anna Poleggi, Dr. Claudia Giannattasio, and Mr. Michele Equestre for genetic and CSF 14-3-3 analyses; Anna Ladogana and Maurizio Pocchiari for critical comments on the manuscript; and Alison Boyd, Genevieve Klug, Samantha Douglas, and Amelia McGlade for their contribution to data management. The authors also thank the Prion Diseases Program, the French National Surveillance Network for Creutzfeldt-Jakob Disease, the National TSE Surveillance Unit, Göttingen, Germany, and the families of CJD patients for their enthusiastic cooperation.
Footnotes
-
Supplemental data at www.neurology.org
Authors' affiliations are listed at the end of the article.
Supported by grants from the Federal Ministry of Education and Research (BMBF 01GI0301 and KZ: 0312720), the Federal Ministry of Health (BMG Az325-4471-02/15), the Robert Koch-Institute through funds of the Federal Ministry of Health (grant 1369-341), the Department of Health (121/7369), the European Union (TSELAB QLK2-CT-2002-81523), the Department of Health and the Scottish Executive Department of Health (The UK National CJD Surveillance Unit), the National Registry of CJD and Related Disorders of the Istituto Superiore di Sanità, Rome, Italy, the Commonwealth Department of Health and Ageing, the Fonds voor Wetenschappelijk Onderzoek, and the Born Bunge Institute.
Disclosure: The authors report no disclosures.
Received October 14, 2008. Accepted in final form March 12, 2009.
REFERENCES
Disputes & Debates: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.